Reliability

APIs, queues, processes, builds, and dashboards kept telling me yes to questions I had not actually asked.

When a public-data source breaks, replacing yesterday’s truth with today’s garbage is not freshness. It is vandalism with a timestamp.

One malformed record should become one visible exception, not a reason to poison or stop an entire daily import.

A public data portal rejected an ordinary collector, and the right response was better failure handling—not a proxy-rotation arms race.

I asked a tiny dashboard for reassurance ten times a second. It answered by manufacturing its own outage.